search

alkaline-ml / pmdarima

A statistical library designed to fill the void in Python's time series analysis capabilities, including the equivalent of R's auto.arima function.
https://www.alkaline-ml.com/pmdarima
MIT License
1.55k stars 228 forks source link

Include confidence intervals in predictions #20

Closed tgsmith61591 closed 5 years ago

tgsmith61591 commented 6 years ago

Description

This issue is related to #18, which referenced lingering include_std_err args. While statsmodels' ARMA model supports this functionality, SARIMAX does not. Opening this for the backlog in case SARIMAX ever catches up to ARMA and we can re-include these parameters in ARIMA.predict.

charlesdrotar commented 6 years ago

Not certain how relevant this is but I found confidence interval support for SARIMAX here.

typpo commented 6 years ago

@tgsmith61591 Thanks so much for creating this project - it's great for people who are new to this area.

Is there any way to derive confidence intervals from the results currently? I'm not sure I can present the results without some sort of uncertainty.

tgsmith61591 commented 6 years ago

Hey there @typpo this is something that I'm targeting for next release. Stay tuned. :-)

tgsmith61591 commented 6 years ago

Just getting around to this. See 8581455 for an initial crack at it. SARIMAX doesn't make this as natural as the ARIMA class, so I had to hack it a bit to make it work.

tgsmith61591 commented 5 years ago

This was merged into dev and will make it into the master branch for release 0.7. Going to leave open until the release.

MaxMcGlinnPoole commented 5 years ago

Very excited for this to make it into master!

RobeeF commented 5 years ago

Thanks !

tgsmith61591 commented 5 years ago

Release 0.7.0 fixes this.